The low volatility anomaly in equity sectors – 10 years later!

Benoit Bellone
is a quant researcher in the Quant Research Group at
BNP Paribas Asset Management in Paris, France.
benoit.bellone@bnpparibas.com, +33 (0)1 58 97 28 42

Raul Leote de Carvalho

is deputy head of the Quant Research Group at
BNP Paribas Asset Management in Paris, France.
raul.leotedecarvalho@bnpparibas.com, +33 (0)1 58 97 21 83

11 August 2020

Keywords: Low Volatility, Low Risk Anomaly, Minimum Variance, Minimum Volatility, Factor
Investing, Equities, Smart Beta, Sectors

JEL Classification: G11, G12, G14, E44

Introduction

At BNP Paribas Asset Management, we are heading for 10 years since the release of our Global Low
Volatility Equity strategy1, which is based on the proprietary research on low risk stocks that we carried
out prior to launching it.

One key result of our research was the evidence that the least volatile stocks from every activity sector
had a higher Sharpe ratio than those of their respective riskier peers. This indicated that the low volatility
anomaly did not appear to be found only in typically less volatile sectors such as utilities, consumer
staples or healthcare. From the point of view of risk-adjusted returns, the least volatile stocks from every
sector of activity appeared to be equally good candidates for a low volatility portfolio designed to deliver
higher risk-adjusted returns and lower volatility than achieved by the market capitalisation index used
to benchmark the strategy.

The strategy we opted for thus invests in the least volatile stocks from all sectors to build a well-
diversified portfolio with below-benchmark volatility and a moderate level of tracking error compared
to those of typical active equity managers.

Almost a decade later, we decided it would be instructive to revisit our research and review the 10 years
of out-of-sample results. Were the least volatile stocks from each sector actually more attractive than
their riskier peers were from the point of view of risk-adjusted returns, as we had found in our initial
research?

As we explain below, the answer is a resounding ‘yes’. If anything, the results were even stronger in
this 10-year out-of-sample period. This runs contrary to the commonly perceived notion that once an
anomaly is discovered it tends to be arbitraged away.

1
The strategy was launched on 31st March 2011

Electronic copy available at: https://ssrn.com/abstract=3697914

The low volatility anomaly

The ‘low volatility anomaly’ refers to the fact that less volatile stocks tend to generate returns that are
higher than would be expected from their level of risk. Robert Haugen and James Heins provided the
first evidence of this anomaly in 1972. They used the history of stock returns to show that, between 1926
and 1969, portfolios investing systematically in the least volatile US stocks would have delivered much
higher returns than could be expected from their low level of risk. Conversely, they showed that
portfolios invested in the most volatile stocks would have significantly disappointed in terms of
performance.

The academic community did not immediately accept their results as they refuted the basic principle in
finance that higher risk should be rewarded with higher return, as advocated by Jack Treynor in 1962
with the Capital Asset Pricing Model. However, this low volatility anomaly has since been confirmed
empirically by many other research studies.

In explaining what started out – and still is referred to – as an anomaly at variance with the basics of
finance, one needs to take into account that the hypotheses used in the formulation of basic financial
theory are statistical simplifications. They have not been, nor will they be, verified in real world
conditions.

First, contrary to prior assumptions, investors do in fact face numerous constraints when investing such
as on the amount of leverage they may use or how much they can rely on short-selling techniques to
arbitrage pricing anomalies.

Second, many investors do not actually seek to maximise absolute returns and reduce volatility. For
example, most professional fund managers are assessed on the relative returns and risk they create
against benchmark indices defined for the purpose.

Third, the assumptions that investors face no transaction costs or taxes, and that markets are perfectly
divisible and perfectly liquid are, as we all know, incorrect.

Fourth, and again contrary to prior assumptions, not all investors have the same investment horizon.
To understand this, one only needs to consider the difference in the time available to a young saver,
relative to his elder peer.

Finally, the idea that information is complete and rationally processed has been challenged by
behavioural theory. Indeed, we know that investors are simply human, so the vast majority is affected
by the same cognitive biases that affect everyone, such as those relating to representativeness,
overconfidence or a preference for lotteries.

All these misconceptions in the assumptions behind basic financial theory can, in ways that have been
discussed by researchers, lead to the low volatility anomaly and the finding that higher risk is not always
compensated with higher return.

Low volatility investing

Low volatility investing has by now become an investment style in its own right with the launch of
numerous low volatility equity funds, in particular since the Global Financial Crisis of 2008. Such funds
invest in low risk stocks and use different risk measures to identify the stocks the fund can buy.

Some of these funds simply aim at being less risky than traditional cap-weighted benchmark indices
while promising higher risk-adjusted returns over the medium to long term. Others aim to outperform

Electronic copy available at: https://ssrn.com/abstract=3697914

these same benchmarks over the medium to long term despite being less risky. However, all these funds
have one thing in common: they assume that the low volatility anomaly will continue to meet their
objectives of higher returns despite lower risk, or just simply of higher risk-adjusted returns than their
respective benchmarks.

On the 31st March 2011, we launched our Global low Volatility Equity strategy. Much like other low
risk equity strategies, ours was tailored to benefit from the low volatility anomaly to deliver on its
objectives. These objectives were defined as outperforming the global market cap-weighted benchmark
over the medium to long term (typically over one or more full business cycles) with less volatile returns
than the benchmark.

The importance of avoiding sector biases

One main difference between our equity low volatility strategy and its peers is the way in which sectors
are considered. Low volatility equity strategies tend to have strong sector biases because the stocks with
the least volatile returns are usually found in sectors such as utilities, healthcare or consumer staples.
This could lead one to believe that the low volatility anomaly is simply a long-term effect arising from
biases towards sectors that are less volatile. This is not the case.

In 2011, we demonstrated empirically that the low volatility anomaly can be observed in every sector of
activity and, as such, it is not a sector effect. The least volatile stocks of every sector have had higher
returns than should be expected from their level of risk, and the most volatile stocks of every sector have
had lower returns than should be expected. These results were eventually updated and published in our
2015 paper, “Low-risk anomaly everywhere: Evidence from equity sectors”, published as a chapter in
the book “Risk-Based and Factor Investing”, ISTE and Elsevier,

Accordingly, the low volatility anomaly is even observed in more volatile sectors such as information
technology or industrials. Given that the least volatile stocks from different sectors have different
absolute levels of volatility, it is important to construct a diversified portfolio invested in the least
volatile stocks of each sector. Simply selecting those stocks with the lowest absolute level of volatility
would result in a non-diversified portfolio, concentrated in stocks from those sectors with the lowest
absolute level of volatility.

Indeed, investing in the least volatile stocks from all sectors adds diversification and delivers higher
risk-adjusted returns than relying solely on a portfolio strongly biased towards the least volatile sectors.
Low volatility strategies that invest across all sectors tend to be more robust in terms of risk-adjusted
returns, even if they may be somewhat more volatile than those that focus only on the least volatile
sectors.

It is perhaps not difficult to understand why creating a permanent sector bias is not a good idea. With
the benefit of hindsight, we know that economically sensitive sectors such as financials, consumer
discretionary, information technology, industrials and materials tend to outperform in the early phase
of the cycle. When activity rebounds, policy is still accommodative, credit and profits start to grow, and
inventories are low and sales improve. More defensive sectors such as healthcare, utilities and energy
tend to underperform in this phase.

Later, in mid-cycle, when activity and profit growth peaks, credit growth is too strong and policy is
neutral, sectors such as information technology and industrials tend to do well, while materials and
utilities usually perform poorly.

Electronic copy available at: https://ssrn.com/abstract=3697914

In turn, late in the cycle, when activity moderates, policy is tight, credit tightens, earnings come under
pressure and inventories grow as sales growth fades, defensive and inflation-resistant sectors such as
materials, consumer staples, healthcare, energy and utilities tend to perform better. In this phase, more
cyclical sectors such as consumer discretionary or information technology typically do poorly.

During the recession phase, with equity markets performing poorly, activity falling, credit drying up,
profits declining, inventories and sales falling and policy easing, consumer staples, utilities and
healthcare tend to do well, while information technology and industrials usually underperform.

While not all cycles are equal, sector rotation has on average been following this pattern for decades.
What makes it difficult to profit from sector rotation is being able to forecast or even nowcast the
changes in the business cycle itself accurately enough.

Low volatility anomaly intra-sectors: global stocks

Ten years after researching this topic, we thought it appropriate to revisit it: how have our research
results held up? Has the low volatility anomaly in each sector been as strong as we found it to be a
decade ago?

Exhibit 1A charts the results of 2011. These were used to develop and promote our global low volatility
strategy. This chart was shown to many investors to help explain why we built our strategy in the way
that we did.

For each sector, we calculated the performance and volatility of two portfolios: one invested in the 10%
least volatile stocks of a given sector and the other invested in the 10% most volatile stocks of the same
sector, picked from the MSCI World index. Both portfolios were rebalanced monthly and the stocks
were grouped into deciles based on their volatility over the preceding three years.

Exhibit 1A shows the Sharpe ratio of such portfolios based on USD net monthly returns. The results
were produced for the first time on 28 January 2011 and were based on a simulation with data from 31
December 1994 through to 31 December 2010.

In exhibit 1B, we show results comparable to those in exhibit 1A, but now calculated on 28 July 2020
using the out-of-sample period from 31 December 2010 through to 30 June 2020.

Electronic copy available at: https://ssrn.com/abstract=3697914

Exhibit 1: Sharpe ratio for the 10% least volatile stocks in each sector and the 10% most volatile stocks
in each sector of the MSCI World index, based on monthly net returns in USD. A: calculated on 28 Jan
2011 based on data from 31 Dec 1994 through 31 Dec 2010. B: calculated on 28 Jul 2020 based on
data from 31 Dec 2010 through 30 Jun 2020. Transaction costs were not included. Source: BNP Paribas
Asset Management, MSCI and Exshare

Both exhibit 1A and 1B show that over both periods, the Sharpe ratio of the least volatile stocks in a
given sector was higher than that of their most volatile peers, almost everywhere. In fact, our 2011
results had one exception, the materials sector. Our out-of-sample results of 2020 had no exceptions.

Another difference is the dispersion of Sharpe ratios across sectors. This is larger in the most recent
period. While our results for the 25-year period used in 2011 span more or less two US business cycles,
our 2020 results are based on a 10-year period that may not yet span one complete cycle. This most
likely explains the larger differences in Sharpe ratio dispersion across sectors in the most recent period.

In exhibit 2, we include the returns in excess of cash and the volatility of each decile portfolio of exhibit
1. What seems clear is that sector dispersion can have a serious effect on the performance of low
volatility strategies strongly biased towards utilities, consumer staples and healthcare.

While the low volatility anomaly is found in each sector in the first set of results running through 2010,
such strategies would have missed out on the performance of the least volatile stocks from the energy
sector. These actually had the highest Sharpe ratio, even though they were not the least volatile of the
investment universe.

In the second set of results, running through 2020, strategies with such a bias would have missed out on
the performance of the least volatile stocks from the information technology sector. These had the
highest returns across the universe and the highest Sharpe ratio.

It is interesting to look at the returns in exhibit 2. In the first period running through 2010, we see that
in consumer staples, energy, financials, industrials, information technology and communication services,
the least volatile stocks delivered higher absolute returns with lower volatility than their most volatile
sector peers did. In the period running through 2020, this is the case for all sectors except for consumer
staples, where there was little difference in returns, and for consumer discretionary.

Electronic copy available at: https://ssrn.com/abstract=3697914

 MSCI World Index universe
31-Dec-1994 through 31-Dec-2010 31-Dec-2010 through 30-Jun-2020
calculated on 28-Jan-2011. calculated on 28-Jul-2020.
Deciles by Sharpe Excess return Volatility Sharpe Excess return Volatility
volatility ratio over cash (%) (%) ratio over cash (%) (%)
Low 0.40 6.3 15.7 0.37 4.5 12.2
Cons. Discr
High 0.22 7.0 31.8 0.34 8.5 25.2
Low 0.77 9.3 12.0 0.88 9.0 10.3
Cons. Staples
High 0.29 6.6 23.0 0.56 9.2 16.4
Low 0.86 13.9 16.2 -0.09 -1.7 17.8
Energy
High 0.48 19.0 39.2 -0.31 -17.8 56.6
Low 0.56 8.3 14.7 0.31 4.3 13.9
Financials
High 0.14 5.8 40.3 -0.16 -4.4 27.2
Low 0.70 10.0 14.4 1.10 12.3 11.2
Health Care
High 0.31 10.7 34.5 0.12 2.7 23.0
Low 0.63 9.0 14.4 0.67 7.6 11.3
Industrials
High 0.15 4.9 33.3 0.13 2.9 22.3
Low 0.57 11.0 19.3 1.13 15.2 13.4
Technology
High 0.18 9.4 51.5 0.18 4.5 24.9
Low 0.54 9.2 16.9 0.60 8.2 13.7
Materials
High 0.54 19.4 36.1 -0.27 -10.4 38.5
Low 0.64 11.2 17.7 0.28 3.2 11.3
Com. Services
High 0.07 3.6 50.0 -0.03 -0.8 25.8
Low 0.70 9.4 13.4 0.67 7.0 10.4
Utilities
High 0.41 14.4 35.3 0.15 3.8 25.7

Exhibit 2: Sharpe ratio, excess returns over cash and volatility for the 10% least volatile stocks in each
sector and the 10% most volatile stocks in each sector of the MSCI World index, based on monthly net
returns in USD. Left: calculated on 28 Jan 2011 based on data from 31 Dec 1994 through 31 Dec 2010.
Right: calculated on 28 Jul 2020 based on data from 31 Dec 2010 through 30 Jun 2020. Transaction
costs were not included. Source: BNP Paribas Asset Management, MSCI and Exshare

Low volatility strategies that diversify by investing in the least volatile stocks of all sectors can more
easily avoid being over-exposed to business cycle rotation in sector returns. They also profit from the
robust finding that the highest risk-adjusted returns, and most often even the highest absolute returns,
can be found in the least volatile stocks of all sectors relative to their respective sector peers.

Low volatility anomaly intra-sectors: Europe

In our 2015 paper, “Low-risk anomaly everywhere: Evidence from equity sectors” mentioned before,
we also found that the results shown in exhibit 1 can in fact be found worldwide. Indeed, in this paper,
we showed similar results for the US, Europe, Japan, Canada, emerging markets, China, Brazil, South
Korea and Taiwan.

In exhibit 3, we revisit the results for Europe calculated over our out-of-sample 10-year period from 31
December 2010 through to 30 June 2020, and we compare them with the results for global stocks.
Because of the smaller number of stocks in the MSCI Europe index (about 450), we now prefer to use
quintiles instead of deciles to provide robust comparisons for both the Europe and World indices.

Exhibit 3 shows that for Europe, the quintile of least volatile stocks in each sector, with the exception
of consumer discretionary, has a higher Sharpe ratio than the quintile with their most volatile sector
peers. For the World index, the use of quintiles instead of deciles does not significantly change the
results shown in exhibit 1. What we find is that, in this 10-year period, the sector dispersion is
comparable for Europe and World stocks. The energy sector and communication services did much less
well than other sectors on a risk-adjusted basis.

Electronic copy available at: https://ssrn.com/abstract=3697914

Exhibit 3: Sharpe ratio for the 20% least volatile stocks in each sector and the 20% most volatile stocks
in each sector, calculated on 28 Jul 2020, based on data from 31 Dec 2010 through 30 Jun 2020. A: for
stocks in the MSCI Europe index, based on monthly net returns in EUR. B: for stocks in the MSCI World
index, based on monthly net returns in USD. Transaction costs were not included. Source: BNP Paribas
Asset Management, MSCI and Exshare

In exhibit 4, we show the Sharpe ratio, returns and volatility of each quintile of stocks behind the results
in exhibit 3. Again, for all sectors with the sole exception of consumer discretionary, the absolute returns
for the quintile of least volatile stocks were higher than for the quintile of their most volatile sector peers.

Another finding is the small differences between the Sharpe ratio, returns and volatility of the quintiles
of European stocks when compared to World stocks in the same sectors. The differences were much
larger when comparing quintiles of different volatility levels in the same sector or comparing results
across sectors in the same region.

Electronic copy available at: https://ssrn.com/abstract=3697914

 MSCI Europe Index universe MSCI World Index universe
31-Dec-2010 through 30-Jun-2020 31-Dec-2010 through 30-Jun-2020
calculated on 28-Jul-2020. calculated on 28-Jul-2020.
Quintiles by Sharpe Excess return Volatility Sharpe Excess return Volatility
volatility ratio over cash (%) ratio over cash (%)
Low 0.42 6.0 14.3 0.44 5.5 12.6
Cons. Discr
High 0.55 13.5 24.5 0.34 8.0 23.5
Low 1.03 11.5 11.2 0.87 8.4 9.7
Cons. Staples
High 0.45 6.6 14.6 0.59 7.8 13.3
Low -0.02 -0.5 18.7 -0.16 -2.9 17.8
Energy
High -0.39 -12.8 32.8 -0.31 -14.3 46.6
Low 0.48 6.5 13.5 0.37 4.7 12.8
Financials
High -0.30 -9.2 30.7 -0.09 -2.1 24.3
Low 0.95 11.3 12.0 1.17 13.4 11.4
Health Care
High 0.65 11.9 18.4 0.48 9.1 18.8
Low 0.75 9.2 12.2 0.59 6.9 11.7
Industrials
High 0.19 3.9 21.1 0.07 1.5 21.0
Low 0.51 9.4 18.5 1.04 14.0 13.5
Technology
High 0.05 1.4 29.9 0.38 8.7 22.7
Low 0.74 10.8 14.6 0.50 7.1 14.1
Materials
High -0.02 -0.5 27.8 -0.19 -6.0 31.7
Low 0.03 0.3 12.3 0.24 2.5 10.8
Com. Services
High -0.08 -1.6 21.2 -0.03 -0.7 20.3
Low 0.70 8.2 11.7 0.53 5.8 10.9
Utilities
High -0.01 -0.3 22.8 0.00 0.0 19.5

Exhibit 4: Sharpe ratio, excess returns over cash and volatility for the 20% least volatile stocks in each
sector and the 20% most volatile stocks in each sector, calculated on 28 Jul 2020, based on data from
31 Dec 2010 through 30 Jun 2020. Left: for stocks in the MSCI Europe index, based on monthly net
returns in EUR. Right: for stocks in the MSCI World index, based on monthly net returns in USD.
Transaction costs were not included. Source: BNP Paribas Asset Management, MSCI and Exshare

From theory to practice

One important point to bear in mind is that portfolios designed to capture the low volatility anomaly are
necessarily constrained in one form or another. Examples of constraints include capping the weight of
stocks, capping portfolio turnover, constraining beta or constraining the tracking error in the case of
benchmarked low volatility strategies.

While important to ensure the portfolios meet investment objectives and to reduce implementation costs
and slippage, constraints can introduce biases that at times may affect performance. The results
presented here serve only as confirmation that the low volatility anomaly is alive and well. The best low
volatility strategies are the ones that find the best compromise to balancing the negative and the positive
impact of constraints when capturing results.

Low volatility investing over short horizons

While our results strongly suggest that the foundations of low volatility investing in the form we
presented 10 years ago remain in place, it is important to include a word about what may be expected
when instead of looking at longer horizons, as done above, the focus turns to the short term.

Predicting the returns of low volatility portfolios over short horizons, e.g. over a month or a quarter, is
not easy, even assuming that portfolio constraints have no impact and that the portfolio is well balanced,
investing in the least volatile stocks of all sectors.

Because of the defensive beta, we can say that low volatility stock portfolios are likely to outperform
the market capitalisation index when market returns are negative, but it is not certain that they will.

Even if the alpha of low volatility stocks is positive on average over the medium and long term, which
explains their higher Sharpe ratios since the alpha is by definition fully uncorrelated with market returns,

Electronic copy available at: https://ssrn.com/abstract=3697914

the occasional negative short-term alpha during a market fall may lead to underperformance. This is
almost inevitable. Similarly, episodes of outperformance of low volatility stock portfolios even when
the market rises, as explained by the positive alpha of low volatility stocks, should not be a surprise.

Takeaways

It is reassuring to see the foundations of our low volatility investment philosophy, based on the research
out more than 10 years ago, being confirmed out-of-sample. Ten years after showing that the low
volatility anomaly in the performance of stocks is a phenomenon that should be considered in each sector
as opposed to on an absolute basis ignoring sectors, we now show that in the last decade, this observation
has held up well, and that if anything, it has become even more valid.

Our results again show how important it is for a low volatility equity portfolio to be diversified and
invested in the least volatile stocks of all sectors. The objective of blindly minimising volatility, resulting
in strong biases towards only a small number of the least volatile sectors, should not be the aim of a low
volatility equity strategy. As an example, in the last 10 years, such strategies would have avoided the
least volatile stocks from the information technology sector. These turned out to have some of the
highest risk-adjusted returns despite not being the least volatile on an absolute basis.

Disclaimer

The views and opinions expressed herein are those of the authors and do not necessarily reflect
the views of BNP Paribas Asset Management, its affiliates or employees.

Electronic copy available at: https://ssrn.com/abstract=3697914